Every prediction backed by peer-reviewed medical research
Key Finding: Protein delays gastric emptying and stimulates incretin hormones (GLP-1 and GIP), resulting in a logarithmic dose-response curve with saturation at approximately 35g protein.
Our Implementation: Michaelis-Menten kinetics with K_m = 18g (half-saturation), maximum glucose reduction of -15 mg/dL at 40-50g protein.
Key Finding: Protein co-ingestion with carbohydrates increases insulin secretion by 32-44% compared to carbohydrates alone, with effects plateauing around 25-30g protein.
Our Implementation: Exponential decay model with saturation point at 25g, maximum insulin amplification of 1.44× (44% increase), resulting in lower glucose peaks through faster clearance.
Key Finding: Meaningful effects on gastric emptying and incretin secretion require a minimum of 10g protein. Below this threshold, protein amounts are insufficient to trigger significant hormonal responses.
Our Implementation: 10g minimum threshold applied to all protein effect calculations. Meals with less than 10g protein receive zero protein-based glucose reduction.
Key Finding: Clinical trial confirming that protein doses ≥30g lower postprandial glucose in both type 1 and type 2 diabetes, with diminishing returns at higher intakes. Validates nonlinear saturation kinetics.
Our Implementation: Reinforces our saturation model with K_m = 18g, confirming that optimal protein doses for glucose reduction range from 25-40g per meal.
Key Finding: Comprehensive review concluding that moderate protein intake (10-30g per meal) optimizes insulin response and glucose control. Higher doses show diminishing benefits.
Our Implementation: Validates our 10g minimum threshold and saturation plateau around 30-40g, supporting our Michaelis-Menten kinetic model for protein effects.
Key Finding: Large cohort study (29,517 adults) established U-shaped relationship between protein intake and diabetes risk. Lowest risk at 12-17% of energy intake, with increased risk at both very low and very high intakes.
Our Implementation: Supports our moderate protein recommendation of 15-20% energy intake for optimal long-term metabolic health and glucose stability.
Key Finding: Review emphasizing that protein's glycemic modulation varies by source (whey, soy, casein). Effects mediated through incretin response and gastric emptying, with all sources showing saturation behavior at high doses.
Our Implementation: Confirms our general protein model applies across sources, though future versions may incorporate protein-type-specific parameters for whey vs. plant proteins.
Key Finding: Comprehensive meta-review supporting ~15-20% energy from protein as optimal for glycemic stability and long-term metabolic health. Aligns with Shanghai cohort findings.
Our Implementation: Validates our recommendation system which suggests protein intakes in the 15-20% energy range for optimal glucose control and metabolic outcomes.
Key Finding: Salivary amylase activity plays a significant role in glucose homeostasis, insulin secretion, and appetite regulation, providing mechanistic insights into individual variations in carbohydrate metabolism.
Our Implementation: This research supports our AMY1 gene copy number approach, demonstrating that salivary amylase levels directly influence how individuals process carbohydrates and regulate blood glucose levels.
Key Finding: Classic work defining white rice GI range (64-93), showing amylose content inversely related to glycemic index. Lower amylose varieties produce higher glucose spikes.
Our Implementation: White rice assigned GI of 73 (medium-high), with recognition that varieties range significantly. Informs our rice-specific calculations in Indian meal contexts.
Key Finding: Compared traditional white rice vs. high-fiber rice in Indian population. High-fiber variety had 23% lower GI (61 vs. 79), demonstrating significant glucose-lowering potential through fiber modification.
Our Implementation: Supports fiber-based GL adjustments in our model. High-fiber rice varieties modeled with lower effective GI for Indian dietary recommendations.
Key Finding: Large meta-analysis of 352,000 participants; each extra daily serving of white rice increased diabetes risk by ~10% due to high GI and GL. Strongest associations in Asian populations.
Our Implementation: Reinforces importance of protein co-ingestion and portion control for rice-heavy meals. Validates our glucose prediction amplification for high-rice Indian diets.
Key Finding: 21-country cohort study; high rice intake (≥450g/day) increased diabetes risk (HR = 1.20), most strongly in South Asia. Dose-response relationship between rice consumption and diabetes incidence.
Our Implementation: Validates targeting Indian/South Asian populations with rice-focused glucose predictions. High portion sizes (>150g) flagged with increased glycemic impact warnings.
Key Finding: Mixed-meal analysis showing that combining white rice with protein (fish, legumes, egg) blunts glucose spikes through delayed gastric emptying and enhanced insulin response. Protein co-ingestion reduces rice GI effectively.
Our Implementation: Core validation for our protein + carb interaction model. Rice meals with ≥15g protein receive significant glucose reduction through combined mechanisms.
Key Finding: Demonstrated that cooled (retrograded) rice yields improved glycemic response via resistant starch formation. Cooling rice after cooking reduces effective GI by 10-15%.
Our Implementation: Future feature consideration: temperature-based GI adjustments for cooled vs. freshly cooked rice in meal predictions.
Key Finding: Food structure intervention study showing that structuring white rice with gellan gum reduces postprandial glycemia. Confirms rice's high baseline GI and potential for food-structure-based interventions.
Our Implementation: Validates high GI assignment for standard white rice (GI 73). Future versions may incorporate structure-modified rice varieties with lower effective GI.
Key Finding: Comprehensive database of glycemic index values for over 2,480 food items, establishing standardized GI measurement protocols and food categorization.
Our Implementation: GI values for all foods in our database, with categorization into Low (<55), Medium (55-69), and High (≥70) GI foods.
Key Finding: Higher-quality carbohydrates—particularly dietary fiber and whole grains—reduce risks of all-cause mortality, cardiovascular disease, type 2 diabetes, stroke, and colorectal cancer. Prospective studies (135 million person-years) showed 15-30% lower all-cause and cardiovascular mortality in highest vs. lowest fiber consumers; whole grains linked to similar benefits.
Clinical Trial Outcomes: Higher fiber intake significantly lowered body weight, systolic blood pressure, and total cholesterol; glycemic index/load effects were smaller or inconsistent, with low-to-very low evidence certainty.
Our Implementation: Our model emphasizes carbohydrate quality through GI/GL adjustments, fiber content integration, and whole grain recognition as protective factors for improved glucose control and long-term metabolic health.
Key Finding: Glycemic Load (GL = GI × carbs / 100) better predicts sustained glucose elevation than GI alone. High GL (>20) leads to sustained elevation, Low GL (<10) has minimal impact.
Our Implementation: GL-based adjustments: High GL (>20) = 1.2× multiplier, Low GL (<10) = 0.8× multiplier, accounting for meal size impact on glucose response.
Key Finding: Prediabetic individuals exhibit insulin overcompensation with reactive hypoglycemia (glucose crashes) 60-120 minutes post-meal, with high-GI foods causing -25% to -35% drops below baseline.
Our Implementation: Extended crash window (10-90 min after peak), crashMultiplier of 1.5, allowing glucose to drop to -35% below baseline for high-GI foods.
Key Finding: Type 2 diabetes patients show 30% higher glucose peaks, slower clearance (fallRate 0.2 vs 0.6), delayed peak timing (1.8× delay), and extended return to baseline (6-8 hours vs 4-5).
Our Implementation: peakMultiplier 1.3, fallRate 0.2, peakTimingDelay 1.8, insulinSensitivity 0.4, extendedClearance 1.8, with 360-minute prediction window.
Key Finding: Healthy individuals maintain glucose peaks of 140-180 mg/dL for high-GI foods, with return to baseline within 2-3 hours. Mild reactive hypoglycemia (-15%) can occur at 60-90 minutes.
Our Implementation: Baseline 95 mg/dL, efficient clearance (fallRate 0.6), mild crash window at 60-90 minutes, maximum -15% crash below baseline.
Key Finding: Physical activity enhances glucose clearance by 10-25% depending on intensity. Light activity (walking) provides 10% improvement, moderate exercise provides 25% improvement, especially beneficial for insulin-resistant individuals.
Our Implementation: Exercise bonus multipliers during glucose clearance: Light activity = 1.1× (10% better), Moderate activity = 1.25× (25% better), applied during falling phase.
Key Finding: Empty stomach (>3 hours fasting) combined with high-GI foods accelerates absorption by 15-25%, with more pronounced effects in prediabetic/Type 2 individuals (25% faster) vs healthy (15% faster).
Our Implementation: Empty stomach + high GI: adjustedPeakTime × 0.75 for prediabetic/type2, × 0.85 for healthy, resulting in faster glucose peaks and earlier crash windows.
Key Finding: Recent food intake (1-3 hours prior) reduces glucose response to subsequent meals by 30-50% due to ongoing insulin secretion and delayed gastric emptying. Very recent meals (<1 hour) reduce response by 50%.
Our Implementation: Recent meal adjustments: 1-3 hours ago = 0.7× factor (30% reduction), <1 hour ago = 0.5× factor (50% reduction), applied to total glucose increase.
Key Finding: Comprehensive scientific dataset combining continuous glucose monitoring (CGM) data with detailed macronutrient intake information, enabling personalized nutrition insights and diet monitoring for improved metabolic health.
Our Implementation: This dataset validates our approach to personalized glucose prediction by providing real-world CGM-macronutrient correlations that inform our algorithm's accuracy in predicting individual responses to foods.
Key Finding: Frequency and consistency of blood glucose monitoring directly correlates with improved glycemic control and better metabolic outcomes. Real-time monitoring enables personalized dietary adjustments.
Our Implementation: Our Food Time Machine leverages continuous monitoring principles to provide real-time glucose predictions, enabling users to make informed food choices based on their personalized responses.
Key Finding: Gaussian Process Regression enables accurate personalized predictions of postprandial glucose responses by modeling individual variations in carbohydrate sensitivity, protein effects, and metabolic state.
Our Implementation: Our predictive algorithm incorporates similar statistical modeling principles to account for individual variability in glucose response, enabling highly personalized predictions that go beyond population averages.
Key Finding: Large-scale genomic studies reveal significant genetic variations that influence individual susceptibility to diabetes and personalized glucose metabolism patterns. Genetic factors account for 30-50% of variance in glucose response.
Our Implementation: Understanding genetic determinants of diabetes risk and glucose sensitivity enables us to provide more accurate personalized predictions based on individual genetic profiles and metabolic predispositions.
Key Finding: Genetic variations in salivary amylase (AMY1 gene) - the enzyme responsible for initial starch digestion - are directly associated with Type 2 diabetes risk and glucose metabolism patterns. Copy number variations in AMY1 influence carbohydrate absorption rates.
Our Implementation: Our AMY1-aware glucose prediction model accounts for individual variations in salivary amylase activity, enabling more accurate predictions for populations with high AMY1 copy numbers (particularly South Asians), where rapid starch digestion significantly impacts glucose response.
Key Finding: Alterations in gut microbiome composition significantly influence glucose metabolism, insulin sensitivity, and Type 2 diabetes risk. Dysbiosis (microbial imbalance) reduces production of short-chain fatty acids and increases intestinal permeability, impairing glucose homeostasis.
Our Implementation: Our personalized glucose prediction model recognizes that individual microbiome composition influences carbohydrate digestion patterns and glucose response variability. Microbiome health status is a key factor in metabolic personalization and long-term dietary success.
Key Finding: Large-scale metagenomic analysis identifies specific microbial strains and species signatures that are strongly associated with Type 2 diabetes risk across multiple populations. Certain bacterial taxa promote glucose dysregulation while others enhance metabolic health.
Our Implementation: Understanding strain-specific microbial signatures enables more granular personalization of glucose predictions based on individual microbiome composition. Users with dysbiotic patterns require different dietary strategies than those with health-promoting microbial communities.
Key Finding: Postprandial glucose responses to identical carbohydrate meals vary dramatically between individuals. Rice is most glucose-elevating overall, but individual responses depend on metabolic phenotype: potato-spikers show insulin resistance with low beta cell function; grape-spikers demonstrate insulin sensitivity; rice-spikers are predominantly Asian. Bread-spikers associate with higher blood pressure.
Mitigator Effectiveness: Preloading with fiber, protein, or fat significantly reduces PPGR in insulin-sensitive individuals but shows minimal effect in insulin-resistant participants—personalization critical.
Multi-Omics Signatures: Continuous glucose monitoring linked to metabolic testing and multi-omics profiling identified insulin-resistance-associated triglyceride profiles, hypertension-associated metabolites, and PPGR-associated microbiome pathways.
Our Implementation: Our model incorporates insulin sensitivity assessments, ethnicity-specific response patterns, and microbiome-aware predictions to personalize glucose forecasts. Mitigator effectiveness scales with individual insulin sensitivity rather than assuming uniform benefits.
Key Finding: A high-protein/low-carbohydrate diet (20% carbs, 30% protein, 50% fat) produced a 40% decrease in 24-hour integrated glucose compared to a high-carb control diet (55% carbs, 15% protein, 30% fat). This dramatic improvement occurred without weight loss—diet composition alone drove glucose control improvements.
Glycemic Control Metrics: Total glycohemoglobin (HbA1c) significantly decreased on the high-protein diet. Postprandial glucose responses were substantially reduced while maintaining stable serum insulin levels, indicating improved insulin sensitivity rather than increased insulin secretion.
Cardiovascular Co-Benefits: Beyond glucose control, the high-protein diet reduced serum triglycerides, increased HDL cholesterol, improved LDL particle size, and lowered blood pressure—addressing multiple type 2 diabetes risk factors.
LoBAG30 Development: This research led to the "Low Biologically Available Glucose" (LoBAG30) diet framework (30:30:40 carb-protein-fat ratio), which became foundational for dietary management of type 2 diabetes and influenced personalized nutrition approaches.
Our Implementation: Our glucose prediction model incorporates protein's dual effects: direct Gannon-effect reduction (-15 mg/dL max via incretin hormones) and van Loon insulin boost (10-20% peak dampening). We recognize protein as both a carbohydrate counter and a metabolic optimizer for improved glucose control.
Try our Food Time Machine to experience research-backed glucose predictions